Slashdot Mirror
Patent 5,893,120 Reduced To Pure Math
An anonymous reader writes "US Patent #5,893,120 has been reduced to mathematical formulae as a demonstration of the oft-ignored fact that there is an equivalence relation between programs and mathematics. You may recognize Patent #5,893,210 as the one over which Google was ordered to pay $5M for infringing due to some code in Linux. It should be interesting to see how legal fiction will deal with this. Will Lambda calculus no longer be 'math'? Or will they just decide to fix the inconsistency and make mathematics patentable?"
This point has been made repeatedly, but nobody cares. The eHarmony patent was shown to be nothing more than linear algebra with particular names assigned to each variable. People have been pointing out the relationship between software and lambda calculus since before most Slashdot users were in high school, but it has had practically no impact on the legality or public opinion of software patents.
Palm trees and 8
The problem is the definition of "implementation." If I patent, say, linear discriminant analysis, and implement it using C++, what did I get a patent on? A C++ implementation? My own personal implementation? No, in the current patent system, I get a patent on any implementation -- or in other words, a patent on the mathematics itself, to within a particular interpretation of the variables and results of the computation.
Palm trees and 8
I've collected various examples of this argument here:
http://en.swpat.org/wiki/Software_is_math
However, you have to remember that this is *not* the end of the discussion (or at best, this will result in this one patent getting invalidated or narrowed).
When faced with "software=math" arguments, judges still argue that *applications* of math to real world problems can be patentable.
What we need is legislation saying that writing, distributing, selling, and using software cannot constitute a patent violation.
Expert in software patents or patent law? Contribute to the ESP wiki!
That's why patents used to require a physical demonstration. Ideas were all well and good, but until you demonstrated a physical device that acted like the patent claimed, you could not get it patented. This was intended to prevent fraudulent devices that looked good on paper but could never actually be manufactured, like perpetual motion machines. It could just as easily prevent software, algorithms and business methods.
It is the central problem of allowing software patents. I think the parent has described why software patents are absurd as concisely as I have ever seen it done. In essence we're allowing the patenting of algorithms, which are at heart pure math. We are allowing the patenting of mathematics, because judges and juries are too fucking incompetent to understand the very basis of computational science. A fraud has been perpetrated on the legal systems of many countries, but rather than throw every single person who has sought a software patent in prison for fraud, we in fact reward them by permitting them to extort money. It's a travesty that very shortly is going to bring the entire industry to its knees. Once you start going after garbage-collected hash tables and refuse to recognize that such techniques are decades old, no one could hope to implement any kind of operating system or virtual machine or, fuck, most interpreters, without risking ending up having their asses sued off.
The world's burning. Moped Jesus spotted on I50. Details at 11.
Sure, software is math. I'll accept the point as stated.
Likewise, books are language. Can books be copyrighted? No one owns language.
Wrong. Words, when put together, form sentences which form paragraphs which form stories, which in turn can be copyrighted. But stories are not paragraphs, paragraphs are not sentences, and sentences are not words so the fact that words cannot be copyrighted by itself is not relevant to stories.
Mathematical formulae however, when put together they're still formulae, merely a longer (and probably more descriptive) one, and mathematical formulae cannot be copyrighted or patented.
There's a big difference between saying "you could probably do that with a computer" and actually doing it with a computer, which is what a software patent would cover.
Except this isn't saying "you could probably do that with a computer", this is saying "you can do that with a computer like this". Allowing software patents but denying math is much like allowing copyright over hexadecimal numbers but disallowing it for decimals: there exist a way to transform one into the other that's formally proven to work for any and all elements of either set, so anything that works for members of one set *must* work for the other and to declare otherwise is to automatically fall into a contradiction. If you'd study Math or CS it'd be much clearer as to how and why, if you must know, but that's kind of outside the context of this post.
No problem is insoluble in all conceivable circumstances.
Nobody is arguing against protecting a specific implementation of an algorithm (although copyright already provides this protection).
The problem is that when you patent an algorithm, you don't just patent that expression of that idea, like you do with a book or a power drill. You patent all the expressions of that idea. It would be like writing a book on some topic, and then owning the rights to all books on that topic until the patent runs out. You own a whole chunk of the language, every possible expression of you idea, not just the particular one you came up with.
Likewise, when you patent an algorithm, you have patented an entire (admittedly fairly confined) branch of mathematics, having to do with expressions of that algorithm. This is distasteful, and, as the current software patent climate has shown, has terrible chilling effects on the software industry as a whole.
There's a difference. Copyright covers only the specific expression. Patents cover the whole idea.
For example, the copyright on Harry Potter covers the story on Harry Potter, and derived works. A patent on Harry Potter OTOH could look like this:
Claims: ...
1. A story about a normal, underprivileged boy who turns out to be special.
2. As 1, where the specialty is that he actually is the son of a magician.
3. As 1, where the boy lives in England.
4. As 2, where the boy himself gets educated in magic.
You see, it would cover a lot of possible books, most of which would have very little relation to Harry Potter. Even worse, it would even apply to books of authors who never heard of Harry Potter (unlikely in case of Harry Potter, but the same would be true for quite obscure books as well). Or imagine that someone else had filed such a patent before, without actually writing such a book, then Harry Potter would not have been a success story, but a nightmare for J. K. Rowling.
Patents have to be much stricter in what they can be applied to because they are much broader in scope.
The Tao of math: The numbers you can count are not the real numbers.
I am an attorney & patent agent and I hold multiple degrees in mathematics and computer science, so I feel fairly competent to speak on this issue.
The claims require a physical implementation (e.g. "An information storage and retrieval system"). No amount of math will produce such a system out of the aether. Nor does thinking about the math or the formulas on pen & paper infringe the patent. The specification makes it clear that "information storage and retrieval system" refers to a computer system. Thus, the patent was not actually reduced to pure mathematics.
The nonpatentability of mathematics refers literally to patenting a formula or algorithm without any useful application, just as chemical elements cannot be patented but a mechanical device made entirely of a single element could be. A claim to a bare formula would be invalid for lack of utility as well as lack of patentable subject matter. Consequently, the Curry-Howard Correspondence has no effect on the patentability of computer-implemented inventions.
Last thing I'd want to do is come up with some whizbang unpatented compression technique, only to find out some MegaMondoCorp has turned it into a multi-billion dollar product leaving me in the dust. on the other hand, I don't want to be sued because my for loop looks too similar to some code sitting in some code base unused for 20 years.
so you want to both have your cake and eat it too?
to be protected yourself but have the ability to go after others?
as it stands it's impossible to tell with certainty if you've violated one of the millions of patents out there as the problem is functionally equivalent to the halting problem.
so you build your lovely wizzbang application and make a profit and someone can swoop in and take it all away over a patent you have not a chance in hell of finding.
All of patent law deals with interpretations, most of which are involve varying degrees of subtly.
The Federal Circuit Court has provided a great deal of well-written guidance. This particularly applies to what is and is not patentable.
The issue of what is and is not patentable is not black and white, such as, “mathematical formulas are not patentable,” or “software is patentable.”
A process that creates something useful and tangible is patentable, whether or not that process involves a calculation. What is not patentable is a “pure” formula that is not tied to something tangible. Data structures are tricky. The newer rules (yes, lots of mistakes were made in the past) are that generalized data structures, such as a table or a linked list, do not count as “tangible.” However, if those data structures are used (critically) to perform useful work, such as to refine steel or to serve up ads on websites, then the ENTIRE process is patentable. Subject, of course to all the other restrictions, such as non-obviousness.
These rules are not really new. They are the same rules that apply to mechanical inventions. For example, you cannot patent a “law of nature,” even it is something complex and nobody else knew about it. You can, however, patent a new device that takes advantage of this law of nature. For example, you cannot patent super-conductivity, but you can patent a useful device that uses super-conductivity.
Even mechanical inventions could be reduced to equations. CAD systems and hardware description languages are such examples. However, these “mathematical” representations have no bearing on the patentability.
Thus, the “deaf ears” referred to are those practitioners in the field who are following well-established law.
You don’t have to like current patent law. Many people don’t. European rules, for example, are different that ours. Note that not liking is distinct from not understanding.
- Registered Patent Agent
I will create a sig when innovation restarts in the U.S.
Copyright covers only the specific expression.
This was long one of the cornerstones of US copyright law. You must have been in a Rip van Winkle nap while that principle was being excised from the law by clever judges over the last few decades. Sadly, you can't even write a completely novel imagining of an older Holden Caulfield's behaviour without falling afoul of the current 'copyright' case law.
Those who would give up essential liberty to purchase a little temporary safety, deserve neither liberty nor safety.